Global Patterns of Freshwater Fish Threatenedness

Exploratory Data Analysis 17.01.24

Justine Carey and Anna-Lisa Dittrich

Introduction

Research question

What are the key anthropogenic or land-related factors which most contribute to freshwater fish threatenedness globally?

Data

  • Species distribution and conservation status

    • IUCN Red List (12,452 freshwater fish species, 2,678 threatened species) (IUCN, n.d.)

  • Catchment boundaries and land use information

  • Connectivity

  • Dams

Data preparation + structure

head(data_withNA)
# A tibble: 6 × 321
    hybas_id continent TotalSppBasin TotalNonDDSppBasin TotalSppThreatBasin
       <dbl> <chr>             <int>              <int>               <int>
1 1060000010 Africa               59                 56                   1
2 1060000100 Africa               81                 75                   2
3 1060000110 Africa               30                 29                   1
4 1060000150 Africa               55                 54                   1
5 1060000160 Africa               60                 59                   1
6 1060000790 Africa               64                 62                  NA
# ℹ 316 more variables: TotalSppExtBasin <int>, ThreatenedRatio_A <dbl>,
#   ThreatenedRatio_B <dbl>, st_Extinct_Total <dbl>, th_Extinct_Total <dbl>,
#   st_Threatened_Total <dbl>, th_Threatened_Total <dbl>, CSI_max <dbl>,
#   CSI_min <dbl>, CSI_wMean <dbl>, CSI_mainStem <dbl>, mainStem_rivOrd <dbl>,
#   mainStem_reachID <dbl>, sum_reaches <dbl>, sum_riverKM <dbl>,
#   area_HyBASIN <dbl>, sum_dams <dbl>, dams_riverKM <dbl>,
#   dams_basinArea <dbl>, WISE_wMean <dbl>, OBJECTID <dbl>, NEXT_DOWN <dbl>, …

Missing values

Analysis

Maps

Hydrosheds downloaded from https://www.hydrosheds.org/products/hydrobasins and combined with the dataset to one Shapefile.

library(dplyr)
library(sf)
hydrosheds<- st_read("Hydrosheds_all.shp")
Reading layer `Hydrosheds_all' from data source 
  `/Users/justinecarey/Library/Mobile Documents/com~apple~CloudDocs/09_study/BOKU/2023-2024 WS/Exploratory Data Analysis in R/Final/Hydrosheds_all.shp' 
  using driver `ESRI Shapefile'
Simple feature collection with 16397 features and 13 fields
Geometry type: MULTIPOLYGON
Dimension:     XY
Bounding box:  xmin: -180 ymin: -55.9875 xmax: 180 ymax: 83.62564
Geodetic CRS:  WGS 84
joined_world <- hydrosheds %>%
  left_join(data_withNA, by=c("HYBAS_ID"="hybas_id"))

Number of species per sub-basin

library(ggplot2)
ggplot()+ 
  theme_void()+ 
  geom_sf(data=joined_world, aes(fill=TotalSppBasin), color=NA)+
  scale_fill_gradient(low="khaki",high="palegreen4", na.value="grey", name="Total species")

Number of threatened species

ggplot()+
  theme_void()+
  geom_sf(data=joined_world, aes(fill=TotalSppThreatBasin), color=NA)+
  scale_fill_gradient(low="dodgerblue2",high="red4", na.value="grey")+
  labs(fill="Total threatened species")

Number of extinct species

joined_world$Extinct <- cut(joined_world$TotalSppExtBasin, 
                            breaks = c(-Inf,c(0,1,2,3),Inf), 
                            labels = c("0","1","2","3","15"),include.lowest=TRUE)
ggplot()+
  theme_void()+
  geom_sf(data=joined_world, aes(fill=Extinct), color=NA)+
  scale_fill_manual(values=c("darkolivegreen3","gold2","orange3","red4"),
                    labels=c("1","2","3","15"))+  labs(fill="Total extinct species")

Correlation matrix

data_withNA %>%
  ggpairs(aes(colour = continent), columns = c(8,14,23)) %>%
  ggplotly(tooltip = c("x", "y", "colour")) %>%
  highlight("plotly_selected")

3D scatterplot

threshold_value <- 2.5
filtered_data <- data_withNA %>% filter(dams_riverKM <= threshold_value)
library(dplyr)
different_obs <- anti_join(data_withNA, filtered_data, by = "hybas_id")
custom_colours <- c("#1f78b4", "#33a02c", "#e31a1c", "#ff7f00", "#6a3d9a","yellow")
plot_ly(filtered_data, x = ~CSI_min, y = ~ThreatenedRatio_B, z = ~dams_riverKM,
                color = ~continent, colors = custom_colours) %>%
  add_markers(marker = list(size = 3))

Principal component analysis (PCA)

PCA_nona <- prcomp(data_noNA[,2:41], center=TRUE, scale.=TRUE)
summary(PCA_nona)
Importance of components:
                          PC1     PC2     PC3    PC4     PC5     PC6     PC7
Standard deviation     3.1654 1.74110 1.67662 1.4792 1.42515 1.35322 1.30937
Proportion of Variance 0.2505 0.07579 0.07028 0.0547 0.05078 0.04578 0.04286
Cumulative Proportion  0.2505 0.32628 0.39656 0.4513 0.50204 0.54782 0.59068
                           PC8    PC9    PC10   PC11    PC12    PC13    PC14
Standard deviation     1.24077 1.1593 1.13093 1.0917 1.05912 1.03677 0.96699
Proportion of Variance 0.03849 0.0336 0.03198 0.0298 0.02804 0.02687 0.02338
Cumulative Proportion  0.62917 0.6628 0.69474 0.7245 0.75258 0.77945 0.80283
                          PC15    PC16    PC17    PC18   PC19    PC20    PC21
Standard deviation     0.95175 0.88344 0.83805 0.81661 0.7721 0.76822 0.61540
Proportion of Variance 0.02265 0.01951 0.01756 0.01667 0.0149 0.01475 0.00947
Cumulative Proportion  0.82547 0.84499 0.86254 0.87922 0.8941 0.90887 0.91834
                          PC22    PC23    PC24    PC25    PC26    PC27    PC28
Standard deviation     0.61292 0.56839 0.56071 0.53587 0.51636 0.48128 0.45242
Proportion of Variance 0.00939 0.00808 0.00786 0.00718 0.00667 0.00579 0.00512
Cumulative Proportion  0.92773 0.93581 0.94367 0.95085 0.95752 0.96331 0.96842
                          PC29    PC30    PC31    PC32   PC33    PC34    PC35
Standard deviation     0.42165 0.41572 0.40581 0.38230 0.3740 0.35599 0.31387
Proportion of Variance 0.00444 0.00432 0.00412 0.00365 0.0035 0.00317 0.00246
Cumulative Proportion  0.97287 0.97719 0.98131 0.98496 0.9885 0.99163 0.99409
                          PC36    PC37    PC38   PC39    PC40
Standard deviation     0.27995 0.24990 0.22596 0.1902 0.09194
Proportion of Variance 0.00196 0.00156 0.00128 0.0009 0.00021
Cumulative Proportion  0.99605 0.99761 0.99888 0.9998 1.00000

PCA - Biplot

library(ggfortify)
autoplot(PCA_nona, data=data_noNA, colour ='ThreatenedRatio_B',
         loadings=TRUE, loadings.colours='blue',
         loadings.label=TRUE, loadings.label.size =3)

PCA - Importance of principal components

library(factoextra)
fviz_eig(PCA_nona)

PCA - Contribution of variables

library(corrplot)
var<-get_pca_var(PCA_nona) #list of matrices containing all results for active variables
corrplot(var$coord, is.corr=FALSE)

Random forest model

Machine learning ensemble algorithm, combines multiple decision trees into robust model

library(randomForest)
library(ggplot2)

set.seed(123)
training_indices <- sample(1:nrow(data_noNA), 0.7 * nrow(data_noNA))

training_data <- data_noNA[training_indices, ]
testing_data <- data_noNA[-training_indices, ]

model <- randomForest(ThreatenedRatio_B ~ ., data = training_data, ntree = 500)
importance(model)
               IncNodePurity
ire_pc_sse         2.0440932
crp_pc_sse         1.8930350
ppd_pk_sav         5.5427870
pop_ct_ssu         4.1981079
ire_pc_use         3.5280902
pst_pc_sse         4.2877228
ppd_pk_uav         6.3457655
pst_pc_use         4.5753773
ero_kh_sav         7.5318707
hft_ix_s09         4.2841666
hft_ix_s93         4.0061011
crp_pc_use         2.6675691
gdp_ud_ssu         2.9039621
ero_kh_uav         7.9507029
pop_ct_usu         3.7881073
hft_ix_u09         3.7826659
urb_pc_sse         0.4743752
gdp_ud_usu         2.6797934
hft_ix_u93         4.2279785
rdd_mk_sav         4.5319660
urb_pc_use         0.5134280
rdd_mk_uav         4.8897086
nli_ix_sav         5.4930418
rev_mc_usu         1.3949523
nli_ix_uav         2.4441400
sum_dams           0.5359298
dams_riverKM       0.7949689
dams_basinArea     0.7624746
riv_tc_ssu         4.9694795
riv_tc_usu         4.6713837
pac_pc_sse         2.8259419
sgr_dk_sav         9.5567625
pac_pc_use         2.9776126
CSI_max            0.1276659
for_pc_use         2.9268159
for_pc_sse         2.7028816
gdp_ud_sav        15.4595145
CSI_mainStem       4.9335657
CSI_min            4.9854327
CSI_wMean          4.1357270
importance_scores <- importance(model)

Random forest model

Discussion

References

Grill, G., B. Lehner, M. Thieme, B. Geenen, D. Tickner, F. Antonelli, S. Babu, et al. 2019. “Mapping the World’s Free-Flowing Rivers.” Nature 569 (7755): 215–21. https://doi.org/10.1038/s41586-019-1111-9.
IUCN. n.d. “The IUCN Red List of Threatened Species.”
Lehner, B., and G Grill. 2013. “Global River Hydrography and Network Routing: Baseline Data and New Approaches to Study the World’s Large River Systems.” Hydrological Processes 27 (15): 21712186.
Mulligan, Mark, Arnout Van Soesbergen, and Leonardo Sáenz. 2020. “GOODD, a Global Dataset of More Than 38,000 Georeferenced Dams.” Scientific Data 7 (1): 31. https://doi.org/10.1038/s41597-020-0362-5.
OpenAI. n.d. “ChatGPT.”
Wang, Jida, Blake A. Walter, Fangfang Yao, Chunqiao Song, Meng Ding, Abu Sayeed Maroof, Jingying Zhu, et al. 2022. “GeoDAR: Georeferenced Global Dams and Reservoirs Dataset for Bridging Attributes and Geolocations.” Earth System Science Data 14 (4): 1869–99. https://doi.org/10.5194/essd-14-1869-2022.
Zhang, Alice Tianbo, and Vincent Xinyi Gu. 2023. “Global Dam Tracker: A Database of More Than 35,000 Dams with Location, Catchment, and Attribute Information.” Scientific Data 10 (1): 111. https://doi.org/10.1038/s41597-023-02008-2.